Re: [boost] Improved Function Object Adapters

From: Mark Rodgers (mark.rodgers_at_[hidden])
Date: 2000-05-17 03:42:06

From: John Maddock <John_Maddock_at_[hidden]
> template <class F>
> struct make_functor_ind;
> 
> template <class T>
> struct make_functor_ind<T*>
> {
>    typedef T type;
>    typedef const T& cast_type;
> };
> 
> template <class R, class A>
> struct make_functor_ind<R (*)(A)>
> {
>    typedef R(*type)(A);
>    typedef R(*cast_type)(A);
> };
> 

Yes, I came to a very similar conclusion, except that I called 
the cast_type param_type and put it into unary_traits.  In fact,
I've shoved the extra indirection into the traits so that users
of the traits (e.g., not1) need not be aware of it.  

I justify param_type as "the type used to pass the function 
object as an argument", and declare my functors' constructors like 
this:

explicit unary_negate(typename unary_traits<Predicate>::param_type x)
   ...

and my helper function like this:

template <class Predicate>
unary_negate<Predicate> not1(const Predicate &pred)
{
   return unary_negate<Predicate>(
     (typename unary_traits<Predicate>::param_type)pred);
}

g++ does not require the cast, and I have been experimenting with ways
to dispense with it for Borland, but have not come up with anything
yet.

I've updated the vault with a new version, but will continue to 
experiment to see if I can improve on it.  Any ideas welcome.

Thanks for your help,
Mark